arxiv Enhancing Q&A with Domain-Specific Fine-Tuning and Iterative Reasoning: A Comparative Study